Asymmetrical pMUTs for Focused Acoustic Pressure by Reinforcement Learning

To increase the energy utilization of a pMUT array, an advanced design scheme for asymmetrical piezoelectric micromachined ultrasonic transducers (pMUTs) has been developed with focused acoustic pressure via the deep deterministic policy gradient (DDPG) algorithm. Three distinctive accomplishments have been achieved in: 1) a highly-efficient interface platform between Python and COMSOL for asymmetry factor (AF) simulations; 2) fast freeform pMUT designs without the initial dataset; and 3) superior designs with increased 34% pressure outputs for potential applications such as contact-less haptics. As such, the proposed design scheme could be applied to other MEMS devices to improve system efficiency.